CSE443 Compilers Dr. Carl Alphonce alphonce@buffalo.edu 343 Davis - - PowerPoint PPT Presentation

CSE443 Compilers

• Dr. Carl Alphonce

alphonce@buffalo.edu 343 Davis Hall

Phases of a compiler

Figure 1.6, page 5 of text

Target machine code generation

7.2.3 Calling Sequence

What happens during a function call?

saved machine status local data actual parameters returned value control link access link temporaries

caller's invocation record

top_sp top

Prior to function call.

7.2.3 Calling Sequence

"Procedure calls are implemented by what are known as calling sequences, which consist of code that allocates an activation record on the stack and enters information into its fields." [p. 436]

saved machine status local data actual parameters returned value control link access link temporaries saved machine status local data actual parameters returned value control link access link temporaries

caller's invocation record callee's invocation record

top_sp top

During function call.

7.2.3 Calling Sequence

" A return sequence is similar code to restore the state of the machine so the calling procedure can continue its execution after the call." [p. 436]

saved machine status local data actual parameters returned value control link access link temporaries … … … returned value … … …

caller's invocation record

After function call.

top_sp top

Caller vs Callee responsibilities

"In general, if a procedure is called from n different points, then the portion of the calling sequence assigned to the caller is generated n

• times. However, the portion assigned

to the callee is generated only once." [p. 436]

Typical calling sequence [p. 437]

"1. The caller evaluates the actual parameters." Recall: formal parameter == parameter actual parameter == argument

saved machine status local data actual parameters returned value control link access link temporaries

caller's invocation record

top_sp top

Prior to function call.

saved machine status local data actual parameters returned value control link access link temporaries actual parameters

caller's invocation record

top_sp top

Caller writes arguments (actual parameters) past the end of its own invocation record.

saved machine status local data actual parameters returned value control link access link temporaries actual parameters returned value

caller's invocation record

top_sp top

Caller knows the

• ffset of the eventual

returned value. When callee returns the caller will look at this location for the returned value.

saved machine status local data actual parameters returned value control link access link temporaries actual parameters returned value control link

caller's invocation record

top_sp top

The caller stores its stack pointer here.

Typical calling sequence [p. 437]

"2. The caller stores a return address and the old value of top_sp into the callee's activation record. … "

saved machine status local data actual parameters returned value control link access link temporaries actual parameters returned value

top_sp

caller's invocation record

top_sp top

The caller stores its stack pointer

• here. When the

callee finishes the stack pointer's value will be reset to this value, thereby restoring the caller's invocation record as the active one (the

• ne on top of the

stack).

Typical calling sequence [p. 437]

"The caller then increments top_sp […] top_sp is moved past the caller's local data and temporaries and the callee's parameters and status fields."

saved machine status

top_sp

Move top_sp saved machine status local data actual parameters returned value control link access link temporaries … actual parameters returned value access link …

caller's invocation record

top_sp

Typical calling sequence [p. 437]

"3. The callee saves the register values and other status information."

saved machine status local data actual parameters returned value control link access link temporaries

PC + machine status

actual parameters returned value access link

caller's invocation record

top top_sp

Write the return address, the current value of the Program Counter (PC), into the saved machine status. When the callee finishes execution will resume with the address pointed to by this saved address. … …

top_sp

PC + machine status top_sp

When control transfers to the callee, the top_sp and top are updated. Callee writes local data and temporaries into its invocation record. saved machine status local data actual parameters returned value control link access link temporaries local data actual parameters returned value access link temporaries

caller's invocation record callee's invocation record

top_sp top

PC + machine status top_sp

If the number of arguments can vary from call to call (e.g. printf) then the caller writes the arguments to the "actual parameters"
 area, as well as information about the number of arguments to the status area saved machine status local data actual parameters returned value control link access link temporaries local data actual parameters returned value access link temporaries

caller's invocation record callee's invocation record

top_sp top

PC + machine status top_sp

If the callee has variable length local data (e.g. local arrays whose size is determined by the value of a parameter) then the arrays are allocated space at the end of the invocation record, and pointers to those arrays are stored in the "locals" block. saved machine status local data actual parameters returned value control link access link temporaries local data actual parameters returned value access link temporaries

caller's invocation record callee's invocation record

top_sp top

Relocatable object code

Compiler produces relocatable object code: addresses are not absolute, but relative to known boundaries (e.g. Stack Pointer, start of record, Program Counter). Linker combines object code files into an executable file, in which static relative addresses are made absolute (in virtual address space). Loader copies contents of executable file into memory and starts execution.

Relocatable object code

Compiler produces relocatable object code: addresses are not absolute, but relative to known boundaries (e.g. Stack Pointer, start of record, Program Counter). Linker combines object code files into an executable file, in which static relative addresses are made absolute (in virtual address space). Loader copies contents of executable file into memory and starts execution.

Leave relative

• ffsets alone during

translation.

Desirable characteristics

• f generated code:

correctness (this is non-negotiable) small execution time small code size small power consumption

Desirable characteristics

• f generated code:

correctness (this is non-negotiable) small execution time small code size small power consumption

Associate costs with each instruction, then "minimize" (lower)

• verall cost, with some

balance since execution time and code size can be in conflict.

Significant tasks of code generator

instruction selection register allocation and assignment instruction ordering

Significant tasks of code generator

instruction selection register allocation and assignment instruction ordering

Which variables are kept in registers?

Significant tasks of code generator

instruction selection register allocation and assignment instruction ordering

Which specific register holds which value?

Significant tasks of code generator

instruction selection register allocation and assignment instruction ordering

E.g. to minimize the number

• f registers needed.

Simple generation strategy vs. code size

If we generate code for each intermediate code instruction in isolation and string the results together the result may include redundant instructions

Small example [p. 509]

Consider: x = y + z This might be translated as: LD R0, y ADD R0, R0, z ST x, R0

<— load the value of y into register R0 <— put into R0 the result of adding R0 and the value of z <— store the value of register R0 to x

Larger example [p. 509]

Consider applying the same template to a larger example: a = b + c d = a + e This might be translated as: LD R0, b ADD R0, R0, c ST a, R0 LD R0, a ADD R0, R0, e ST d, R0

Larger example [p. 509]

Consider applying the same template to a larger example: a = b + c d = a + e This might be translated as: LD R0, b ADD R0, R0, c ST a, R0 LD R0, a ADD R0, R0, e ST d, R0

This instruction is redundant: it is loading into R0 the value that is already there.